Closed esalagran closed 2 years ago
Hi,
You can select the subset of outputs you are interested in, following this file https://dl.fbaipublicfiles.com/convnext/label_to_words.json
Closing this for now. Feel free to reopen if you have questions
您好,请问我想测试自己的数据集,应该怎么修改代码?
I understand that convNeXt with the 22k imagenet is aimed to be a zero-shot algorithm. Is there any way to evaluate the model only on the classes I am interested to?